Adaptive M-QAM signaling for dynamic spectrum access

In this paper we consider an OFDM multiuser system with K orthogonal subcarriers. We assume each user uses M-QAM modulation in each subcarrier. The modulation must be adaptive to cope with propagation changes and to meet the bit error rate requirement of the user. We maximize the sum of information bits per user selfishly, and compare the performance with other power allocation algorithms. Surprisingly, measuring rate by constellation size instead of Shannon's capacity yields a power allocation algorithm that differ widely from previously known power allocation algorithms, such as the iterative water filling algorithm (IWFA). This is because we only accept bit error rates better than 10^-3 and the rate achieved with modulation can only be discrete values, whereas other algorithms based on Shannon's capacity can achieve any rate.